Spectral regression based subspace learning for face recognition

摘要

The current difficulties in face recognition are the computing complexity under the uncontrolled environment. This paper proposes a face recognition algorithm based on spectral regression subspace learning with local binary pattern (LBP) features. Firstly, Gaussian filtering and down-sampling are used to build the image LBP pyramid, from which LBP operator is adopted to extract the LBP features of each sub-image. Then, the multi-scale LBP histogram features are fed as the input of the spectral regression to extract the eigenvectors in the projection face subspace. Experiments results indicated that the multi-scale LBP-SR features are rotation invariance and translation invariance. The spectral regression subspace learning with LBP has better performance in the complex background with fast recognition speed, which can be used for real-time video surveillance.